Parallel multiscale context-based edge-preserving optical flow estimation with occlusion detection
作者：
Highlights：
• We construct a parallel multiscale context network for occlusion detection, which extracts multiscale context information to refine the occlusion boundaries.
• We combine the PMC network with a context network to establish an occlusion detection module and incorporate it into a pyramid, warping, and cost volume network to construct an edge-preserving optical flow model.
• We exploit a novel loss function by integrating an edge loss with an EPE-based loss and a binary cross-entropy loss. The proposed loss function supervises the network to estimate flow field and occlusions simultaneously.
摘要
•We construct a parallel multiscale context network for occlusion detection, which extracts multiscale context information to refine the occlusion boundaries.•We combine the PMC network with a context network to establish an occlusion detection module and incorporate it into a pyramid, warping, and cost volume network to construct an edge-preserving optical flow model.•We exploit a novel loss function by integrating an edge loss with an EPE-based loss and a binary cross-entropy loss. The proposed loss function supervises the network to estimate flow field and occlusions simultaneously.
论文关键词：Optical flow,Occlusion detection,Edge-preserving,Parallel multiscale context,Convolutional neural network
论文评审过程：Received 15 June 2021, Revised 12 October 2021, Accepted 2 November 2021, Available online 17 November 2021, Version of Record 23 November 2021.
论文官网地址：https://doi.org/10.1016/j.image.2021.116560
